ESP project 3A amp at +/-50V

I need a cheap amplifier for bi-amping purposes. So the first amplifier I thought of is actually the ESP project 3A amp. My Aleph will run the tweeters.
However, since I already have two 500VA 2*38V transformers I would like to use them.
But a 2*38V transformer will create +/-50V supply voltage a bit too high..
So I'm thinking of using Sanken 2SC2922/2SA1302 200W as output transistors.

But 50V rails and a 4Ohm speaker will cause a peak dissipation of 156,25W (25*25)/4..
This means that there is no, or very little safety margin if the load is inductive, which I assume a typical speaker is..
Is this OK?
I don't want to blow up the output stage as this will probably destroy my speakers too.

The alternative would be to use two pairs of 2SC2922/2SA1302 per channel. Or buy a new transformer.

/Freddie
 
Thanks for the reply,

You're right!
My mistake, I meant 2SC2922/2SA1216..

Yes, I could use a detector circuit which disconnect the speaker if they see DC.
But I want to now how close to the margins I could drive the transistors, as I have no experience in amplifier design.

Maybe I should also mention that I'm not going to push the amp very hard.. And the heatsink is about 0,34K/W one per channel.

/Freddie
 
YW, I recall reading about SOA, safe operating area for bipolar transitors. It is usually displayed as a graph showing the product of all possible currents at all possible voltages. The area under the curve was the graphical representation.

But, I think that if you just try it and be sure to use the 2 amp fuses in line with the speakers, you should be ok. I don't have my Japaneese spec manual handy; what are the current ratings of those output transistors? If they handle at least 10 amps you should be OK since transistors often can peak out at twice their rated current.
 
max pdiss in the output stage = 2* Vcc*Vcc/ pi*pi*cos (phi) *Z
(assuming class B)

vcc = 50
z= 4

for a purely resistive load pdiss turns out to be about 127 W, so 150W is probably more reasonable

delta-T for the transistors is (delta-T is the temperature rise from the ambient temperature)
150 *( rtj-c + rtc-s + rt s-a) where
rtj-c = 125 /200/2 = 0,3125 C/W (max temp 150 C, P =200W, two transistors )
rtc-s = 0,5/2 = 0.25 C/W (assuming isolation pads)
rts-a = 0,34 C/W

=> delta -T = 135 C which means that with 25 C in the room
your transistors are gooners :( (155 C is to much)
Run doubble pairs... gives delta-T = 93 C which has a good safety margin.

This is of course at maximum loss situatuin which happends at roughly half maximum output power for a class AB amplifier. So just by keeping the volume down might not help to save the transistors.

/rickard
 
I like how you did the calculations. It is pretty close. I have seen lots of name brand amps, and I would say that in general, those with no no load power supply voltages up to about +/- 55 volts only use one set of those type of output transistors.

The manufacturers don't make the power supplies as stiff for most amps in equipment like stereo receivers. Therefore, the outputs don't have to drop as much voltage under a load.

I kind of like that approach since it provides some dynamic headroom yet reduces output transistor disipation. I tend to try to get by with the minimalist approach for my projects in regards to parts counts.

Around 1980, when Dan Meyer's Tiger amp was several years old, I had aquired the circuit boards and output assemblies. I was new to amps back then, and after experimenting with them a lot, the outputs blew out. I finally replaced them with 2N3055's and MJ2955's even though the transformer was about 68 V center tapped. But it was only about a 300VA transformer for both channels, and the filter caps were only 2200 microF. They charged up to almost 48 vollts when the amp was idling. Another factor someone else discussed earlier is the part manufacturer. Some manufacturers make parts that actually exceed the specs by a good margin.

But your transformers are rather big and you were considering using large filter cap values. So two pairs of output transistors should do well.